Conditional Abstraction Trees for Sample-Efficient Reinforcement Learning

31 Aug 2023

In many real-world problems, the learning agent needs to learn a problem’s abstractions and solution simultaneously. However, most such abstractions need to be designed and refined by hand for different problems and domains of application.

Who Needs to Know? Minimal Knowledge for Optimal Coordination

17 Aug 2023

It is often crucial to have information about one’s collaborators. However, not every feature of collaborators is strategically relevant.

SMCP3: Sequential Monte Carlo with Probabilistic Program Proposals

21 Jul 2023

This paper introduces SMCP3, a new family of sequential Bayesian inference algorithms.

Dealing with expert bias in collective decision-making

12 Jul 2023

Quite some real-world problems can be formulated as decision-making problems wherein one must repeatedly make an appropriate choice from a set of alternatives.

Who Needs to Know? Minimal Knowledge for Optimal Coordination

07 Jul 2023

To optimally coordinate with others in cooperative games, it is often crucial to have information about one’s collaborators. However, not every feature of collaborators is strategically relevant.

Seventh annual CHAI Workshop

20 Jun 2023

From the 16th to the 18th of June, CHAI held its 7th annual workshop at Asilomar Conference Grounds in Pacific Grove, CA.

Harms from Increasingly Agentic Algorithmic Systems

29 May 2023

CHAI Micah Carroll and former CHAI interns Dmitrii Krasheninnikov, Lauro Langosco, and Yawen Duan published a paper titled Harms from Increasingly Agentic Algorithmic Systems for FAccT 2023.

Dealing with expert bias in collective decision-making

02 May 2023

In their paper published in the AI journal, CHAI Tom Lenaerts and Axel Abels argue that quite some real-world problems can be formulated as decision-making problems wherein one must repeatedly make an appropriate choice from a set of alternatives.

Keynote at The Value Connection workshop

11 Apr 2023

On the 1st of April, CHAI’s Brian Christian gave an opening keynote address at “The Value Connection Workshop” in Amsterdam, an interdisciplinary workshop on human values, hosted by TU Delft and sponsored by the Royal Netherlands Academy of Arts and Sciences and the Delft Design for Values Institute.

AI has much to offer humanity. It could also wreak terrible harm. It must be controlled

05 Apr 2023

On the 2nd of April 2023, Stuart Russell published an article in The Guardian making the point that systems with abilities exceeding human capacity have been let loose.